Your playbook, applied to every clause
Specialist agents review a contract against your own rulebook, not a generic template. Every finding cites the text it came from, replacement language is drafted where a position is missing, and anything the system is unsure of goes to a lawyer rather than being decided quietly.
This is not a proposal. It is Paralegent AI, our own product, in production today. Here is how it was built.
23
agents in the review pipeline, scoring then analysing
5–10 min
for a review, inside Word, on a real contract
0
findings without the clause they came from
Playbook-driven review, inside Word
Five costs of reviewing contracts by hand
We do not publish a saving, because it depends entirely on your volume and your rates. Each card names the arithmetic instead, so you can run it with your own numbers.
Hours spent on each contract
Reviews are measured in hours of a lawyer's time, and the queue never gets shorter.
Count your contracts per month against your loaded legal cost.
Critical risks slip through
A tired reader on contract forty misses what a fresh one catches.
One missed clause can cost more than years of review time.
Inconsistent reviews
Different reviewers apply different standards to the same clause.
Re-review work nobody budgets for.
A bottlenecked legal department
Legal becomes the slowdown in every deal, so business teams start routing around it.
Sales cycles stretch while contracts wait.
Expensive outside counsel
Routine agreements go to external lawyers at partner rates because there is no capacity inside.
Your own outside-counsel line already knows this number.
Lawyers review findings, not raw text
The judgement stays with the lawyer. What goes away is the reading of forty pages to find the six that matter.
Before
- Manual review measured in hours of a lawyer's time
- Risky clauses spotted only if the reader is fresh
- Different reviewers applying different standards
- Legal as the bottleneck on every deal
- Routine contracts sent out at partner rates
After
- A review in minutes, with the lawyer reading findings rather than pages
- Risky clauses flagged and replacement language already drafted
- The same playbook applied to every contract, every time
- Standard contracts self-served, with legal setting the rules
- Outside counsel reserved for the genuinely hard cases
Score first, then route to a specialist
Twelve scoring agents run in parallel, one per legal category, and route only what matters to eleven specialist analysts.
Running every specialist against every contract would be simpler to build and far more expensive to run: smart routing cut LLM calls by 75% compared with the brute-force version. That is why a review costs what it costs, and it is the kind of decision that only shows up once a system has real contracts going through it.
Your playbook is the input
Firms that already have a written playbook start faster, because the hardest part of this project is deciding what your positions actually are — and you have already done it.
Day 1
Upload your playbook
We digitise your standards and preferred positions — the document that already exists, parsed into terms the system can apply.
Weeks 1–2
Tune to your positions
The rulebook is configured to your specific requirements and risk tolerance, and tested against contracts you have already reviewed.
Week 3
Integrate the workflow
Connected to the tools your team already works in — Word, your contract system, email — so nothing new has to be learned.
Week 4
Go live
Reviews start running against real inbound contracts, with a lawyer approving every output.
Contract review is one part of it. For the wider picture, obligation extraction, renewal and notice-date intelligence, record-versus-document reconciliation and legal document intelligence beyond contracts, see AI for legal operations on Dynamics 365 and SAP.
Deep dives from our engineering team
How do you review a contract against your own legal playbook with AI?
A contract review system that reads your own playbook is four services and three queues. This is the whole pipeline — what each stage costs, where it degrades, and the one design decision the architecture is built around.
Read more8 min read
Twelve scorers, eleven analysts — what happens to the category with no specialist?
A contract-review pipeline scores every chunk across twelve legal categories and routes it to eleven specialists. The missing twelfth is not an oversight — it is the control signal the whole cost model depends on.
Read more8 min read
How do you stop eleven agents reporting the same clause eleven times?
Parallel specialist agents all read the same text, so they all report it. Deduplicating across agents is not tidying — it is what decides whether the output is usable, and it is a different problem from deduplicating messages.
Read more6 min read
Smart Category Routing for Contract Review
A focused application of the LLMOps routing pattern to legal contract analysis — the analyst-selection logic that ships fewer clauses to fewer agents and finishes a 3,300-call review in 154 seconds.
Read more6 min readWhere should retry live when your worker reads from a queue?
A queue worker can retry in-process or let the queue redeliver. Doing both is correct — and the two tiers multiply rather than add, which is the cost nobody budgets for.
Read more6 min readWhat stops a degraded model from taking the whole queue with it?
Retry is how a degraded model provider becomes an outage. A circuit breaker fails fast instead — and in a container fleet it protects far less than you think, because every replica has its own.
Read more6 min readIs partial failure an outage, or a setting?
When three of twelve agents fail, does the job fail? Making that a configured threshold rather than an exception is the right call at scale — and it produces results that carry no record of how complete they are.
Read more6 min readDeterministic parsing or a model — when is the cheapest call the one you never make?
The same pipeline ingests a spreadsheet with no model calls and a PDF with dozens. The format your customer sends decides your unit economics — which makes the upload form a cost lever most teams never touch.
Read more7 min read
Send us a contract you have already reviewed
This is our AI engineering practice
It is real work and it is where our four products came from. But what Cognilium leads with is narrower: optimization apps that run in tandem with Microsoft Dynamics 365, computing the decisions the ERP records but does not derive — the optimal price, the optimal pick path, the optimal stock level. See the optimization apps · How we build inside the ERP.